1 / 18

Overview

Overview. General View of DAVINCI development Developing non-XDM algorithms. What is DAVINCI. Typical application runtime. The development procedure. Developing non-XDM algorithms. Application range. Development overview.

nasia
Download Presentation

Overview

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Overview • General View of DAVINCI development • Developing non-XDM algorithms

  2. What is DAVINCI

  3. Typical application runtime

  4. The development procedure

  5. Developing non-XDM algorithms • Application range

  6. Development overview • Codec Engine provides an easy-to-use application interface-VISA-for multi-media developers to integrate and deploy compliant codecs. CE uses the codec’s xDM interface to call specific processing and control functions. • Developing non-XDM is extend CE by adding new VISA-like APIs for remotely accessing xDAIS algorithms that do not implement the xDM interface.

  7. 1.The GPP-side ap-plication makes an algorithm MOD_pro-cess() call. 2.CE forwards this call and its argume-nts to the GPP-side algorithm stub. 3.The stub places the arguments in an inter-CPU message and replaces all GPP-side (virtual address) pointer values with DSP-side (physical address) values. This called “marshalling” the arguments. 4.CE delivers the message to the DSP-side algorithm skeleton. 5.The skeleton unmarshals the arguments and calls the actual xDAIS algo-rithm’s process() function. 6.On the way back, the skeleton marshals any return arguments, places th-em in a message, and the stub unmarshals them for the application.

  8. Tasks (The core xDAIS algorithm interface is alre-ady there.) • Create CE extension layer. • Custom stub interface • Custom skeleton interface • Perform CE packaging and configuration (Codec Server & Codec Engine engineers’ job)

  9. Create CE extension layer • Design a new application interface, including the following core functions: ROTATE_create(), ROTATE_delete(), ROTATE_process()

  10. ROTATE_Handle ROTATE_create(Engine_Handle server, String name, ROTATE_Params *params) { ROTATE_Handle visa; static Bool curInit = FALSE; if (params == NULL) { params = &ROTATE_PARAMS; } visa = VISA_create(server, name, (IALG_Params *)params, sizeof (_ROTATE_Msg), "ti.sdo.apps.extensions.rotate.IROTATE"); return (visa); }

  11. Void ROTATE_delete(ROTATE_Handle handle) { VISA_delete(handle); }

  12. XDAS_Int32 ROTATE_process(ROTATE_Handle handle, char inBuf[], char outBuf[], XDAS_Int32 bufSize, XDAS_Int16 cosine, XDAS_Int16 sine) { Int32 retVal = ROTATE_EFAIL; ROTATE_InArgs inArgs; inArgs.bufSize = bufSize; inArgs.cosine = cosine; inArgs.sine = sine; if (handle) { IROTATE_ADAPT_Fxns *fxns = (IROTATE_ADAPT_Fxns *)VISA_getAlgFxns((VISA_Handle)handle); IROTATE_Handle alg = VISA_getAlgHandle((VISA_Handle)handle); if (fxns && (alg != NULL)) { VISA_enter((VISA_Handle)handle); retVal = fxns->process(alg, (XDAS_Int8 *)inBuf, (XDAS_Int8 *)outBuf, &inArgs); VISA_exit((VISA_Handle)handle); } } return (retVal); }

  13. Create CE extension layer • Develop CE stubs • Define a new VISA “message” type used for marshalling and unmarshalling call • Marshal the call and its arguments using the VISA “message” • Dispatch the call to the server • Unmarshal the returned arguments • Free any required VISA message • Return the results of remote computation

  14. /* ======== rotate_stubs.c ========*/ _ROTATE_Msg *msg; if ((msg = (ROTATE_Msg *)VISA_allocMsg(visa)) == NULL) { return (ROTATE_ERUNTIME); } //Create the “message” msg->visa.cmd = _ROTATE_CPROCESS;//Set the command ID msg->cmd.process.inBuf = (XDAS_Int8 *) Memory_getBufferPhysicalAddress(inBuf, inArgs->bufSize, NULL); if (msg->cmd.process.inBuf == NULL) { retVal = ROTATE_ERUNTIME; }//Marshalling:Translate the address references to server’s address space msg->cmd.process.outBuf = (XDAS_Int8 *) Memory_getBufferVirtualAddress(outBuf, inArgs->bufSize, NULL); if (msg->cmd.process.outBuf == NULL) { retVal = ROTATE_ERUNTIME; }//Unmarshalling msg->cmd.process.inArgs = *inArgs; retVal = VISA_call(visa, (VISA_Msg *)&msg); exit: VISA_freeMsg(visa, (VISA_Msg)msg); return (retVal);//return the rusults of remote computation

  15. Create CE extension layer • Develop skeletons for CE extension • Select the dispatch function by parsing the command ID in the message • Unmarshalling “processing” function arguments • Call the core xDAIS algorithm interface function • Marshal return arguments into the reply message • Send the reply message back to the caller

  16. /* ======== rotate_skel.c ========*/ static VISA_Status call(VISA_Handle visaHandle, VISA_Msg visaMsg) { _ROTATE_Msg *msg = (_ROTATE_Msg *)visaMsg; ROTATE_Handle handle = (ROTATE_Handle)visaHandle; switch (msg->visa.cmd) { case _ROTATE_CPROCESS: { inBuf = msg->cmd.process.inBuf; outBuf = msg->cmd.process.outBuf; Memory_cacheInv(inBuf, msg->cmd.process.inArgs.bufSize); /* make the process call */ msg->visa.status = ROTATE_process(handle, (char *)inBuf, (char *)outBuf, msg->cmd.process.inArgs.bufSize, msg->cmd.process.inArgs.cosine, msg->cmd.process.inArgs.sine); Memory_cacheWbInv(outBuf, msg->cmd.process.inArgs.bufSize); break; } default: { msg->visa.status = VISA_EFAIL; break; } } return (VISA_EOK); }

  17. Packaging the extension Compile the packaging file: package.xdc, package.xs, package.bld, IROTATE.xdc, makefile Build the extension package

  18. Advice and Criticism are warmly welcomed!

More Related